CEO Confidence in AI Slips as Chip Shortages and Water-Free Data Centers Reshape IT | TSG Ep. 1003

Hey everyone. It's Thursday and we're live. I think we're live.

Uh, if you're watching this at about 10 o'clock eastern time, then we're live. Uh, welcome to our Textron gang chauffeur today. We've got some great gang folks to introduce you to, to participate.

We've got, uh, from ALP in, that's, uh, Pacific Northwest. My friend Fred Wilmot. We've got Gina Rosenthal.

Is that a new hand dude, Gina, or just a maybe Your camera's gotten a lot better. I, I'm, I'm really, I'm seeing you out there, lady. Thank you.

Appreciate it. And then also joining us from out Westley. We've got Jon Swartz, and of course we've got the Dean, Mike Vizard, Dean ard, um, welcome gang members.

It's certainly interesting times. You gotta love what's going on with Davos and, and the rest of the world. Just no shortage of things for us to talk about.

But, um, you know, while AI was probably the darling of Davos, it's maybe not the darling of CEOs these days. I don't know. That's a sy Mike Sy, what is this about?

Well, I'm gonna let John explain it, but it seems like maybe the bloom is off the aios a little bit. 'cause CEOs are trying to maybe come to terms with the fact that, well, the things that employees have been saying for a while might be true, and it doesn't have that impact on the top line or the bottom line. But John, walk us through this a little bit.

Yes. So there was a PWC global CEO survey, and it's a pretty significant one. It was about 4,500 leaders across nearly a hundred countries.

And they found in their study that more than half, about 56% say their companies have no, have seen, no financial return from AI investments So far. Uh, even more telling is 12% said they achieve both lower costs. Uh, only 12% said they have achieved both lower cost and higher revenue.

So they're, in a weird way, this kind of mirrors to a more reasonable degree. This MIT study that came out last summer, which was, I think really flawed, but the same, the gist was, is that AI is not returning the investment despite all the spending, and that the progress is slow. Um, to add on to that and send, something that Mike shared with me late yesterday was s and p Global ratings also released two reports and, uh, showing the same similar concerns.

So we have these studies and, and, and they kind of move towards this, this conclusion that things having caught on as quickly as as as we expected. Now, there's going to be yet another survey coming out next week. I can't tell you who it's from because I'm under embargo, but basically it says something a little bit differently.

They're saying, G Gen generative AI adoptions increased 44% from a year ago, and half of the response instead, they're moving ahead with the Gen ai. So there's kind of a little bit of a more optimistic point of view. And then finally yesterday, and we'll, we'll talk about more about this tomorrow, so I'm not gonna talk too much about it now, but Jensen Wong of of Nvidia presented this idea that, uh, the, uh, trade associations and the, the people who work at is plumbing electricians.

They're, they're gonna benefit greatly from ai. So we've got all these different viewpoints. It's kind of like the Rashman movie from, uh, awa, right?

Depending on who the person is and what their motivation is, we have a different point of view of where things are going. So there's seems to be a mishmash, but I I, I do give creams to PWC study because they are a, uh, credible organization. And they did talk to a fair number of people, a lot of people actually across the world.

And I think that's where we are now. I mean, it's up to debate. We might see more reports that go counter to this, but I do think there really is a, a debate is live about how far and how deep this pen penetration and useful investment and return on investment that AI has reached.

It seems to me at least that everybody's using AI and they're getting some sort of benefit out of it. It's just not enough to move the needle apparently. And I think CEOs are figuring out it may be a little bit longer before that needle really moves.

It will eventually move, I think. But, uh, it doesn't feel like employees are particularly well trained on how to take advantage of it. It doesn't feel like the tools themselves are all that easy.

And it's still the first inning of all of this for, yeah. Yeah. It's, it is like a trial and error.

I mean, this is, we, we keep hearing that we're in the very, very early innings of the AI adoption. And I, I'm positive that it will accelerate this year. But even these things take time and it's a major, uh, overhaul of how companies do business in the employee are in the needle.

They're, they're under a lot of pressure. Learn fast. So look, here's, here's a shimmy take on this.

You didn't find your return on investor, your ROA on ai dig harder. Get me 11,000 more votes await. Someone said that.

But guys, here's, here's, here's the real deal, right? And, um, this, this, honestly shimmy's take on it. The fact of the matter is agentic AI is a lot more myth than reality today.

As we've said before, most of these early on AI agents suck. They don't do what you want 'em to do, and they, they claim to do this, and then you see how they do it, and it leaves a lot to be desired. So the amount of work and effort involved to get these agents to do something is just too great.

Yet we're, we're still perfecting that. When you look at generative ai, it's great. It does so many cool things.

Every single one of us here, I bet, is using it and finding it in some ways, automagical, some ways not, but it, it, the kinds of things it's doing is not easily reflected in your ROI bottom lines, right? It's helping me write an email. It's helping me with my articles, it's helping me with some research.

But where does that translate to the bottom line? So there's, that's number one. Number two is this, Mike, and you touched on it a little bit specifically when it comes to generative ai.

And I, I'm talking now as the CEO of Textron. I have hammered entire people, Hey, you're not gonna lose your job to ai. You're gonna lose your job to someone who uses AI better than you.

But consistently, when I see how our people are using ai, I'm disappointed. And I'm talking about generative ai. Now, they don't understand how to craft a prompt to get them what you need.

They're, they're, they're tasks that they're giving AI are framed in almost third grade kind of questions. You, you know, it's only as good a tool as the tool user as the craftsman using it. And, you know, don't blame, don't blame the tool here.

Blame the guy use or the gal or the person using the tool. I, I, I, you know, I just went through this last night yet again where I took what one of our people did and said, this isn't really good, you know, using ai. And I, I made my own prompt.

I ran it again for them and said, here, this is a lot better. This is what's useful. This is what you should be doing.

'cause otherwise, someone who does know how to do it is gonna take your job. I didn't say that to that, but that's the truth. And I, I think that's our biggest issue right now, is when we talk about generative ai, we gotta move beyond the people who write a simple sentence of seven word sentence and expect AI to return gold.

And then we've gotta make agentic AI actually work and work easy enough to get through the autonomous task we wanted to do Well. But let's look at the population then and say, okay, what percentage of the population has the skills to craft a really good prompt or the expertise or the knowledge? And you come up with numbers that are gonna be well below 20% because They, but isn't that really, isn't that really a reflection of the American education system that people can't read it?

Right? And if you are going to take a shot at the American system, I'm not gonna sit here for this. I'm not go with the rest of my Delta house guys outta here, dude.

That's the, So the issue become asking AI to make up for the fact that that issue exists is unreasonable expectation. So therefore, we gotta figure out how are we gonna get that 80% to craft a better question. And some people say, you know, the true sign of intelligence is not the ability to know everything and the ability to frame the question in the first place.

But Fred, I, That's exactly it. I think, Oh, go ahead. Let me, let me frame that question.

Uh, as you said, Mike, uh, let's ask that question. How long did it take to adopt email? How long did it take to adopt ethernet?

How long did it take to adopt the internet at large in order to operate businesses longer than two or three years across the board Much longer. And in, in that process, what was fundamentally important about the adoption rate was the level of education. It's unfair to suggest that every person with full equality of their educational experience, shimmy, or their technical competency to suggest they should have the equitable, equitable learning strategy of how to become competent or proficient in ways to utilize AI for the business.

And the business owns some accountability here, right? With that process. And I think, I like the way that you suggested that folks need to understand their role is to be able to use AI effectively, not be replaced by ai.

But fundamentally, technology shifts don't happen overnight. CEOs aren't going to see a dramatic increase in their return on investment. But I think that same survey also demands that if you sit still while this process happens, you're gonna be at least five points behind.

And even with the uncertainty in the market, right? When thing's always true, when you invest in your people, better outcomes happen. So the chaos around this thought process of whether it, whether or not AI serves your bottom line, a better question might be, do your people serve improvement in your bottom line?

If so, invest in your people. Allow them to bring things to market. Innovation's a real thing.

And I think that supports the, the ideology of what AI is supposed to bring to the business. Anyway, I have two things on that, right? So I, I agree with what you're saying there.

I also think that we have to look at the workforce at huge amount of the experience. Workflow Force has been let go to fund ai. Um, so that's number one.

But you look at the people that are left, they may not have the vocabulary to form the questions and to create the prompts because the people they would've relied on to help them get that vocabulary are now gone. And they're having to figure it out from Grand Zero on their own. And this is not just ai that, right?

So we're talking specifically here about generative ai. One thing I found interesting from the article is that, um, PWC also thinks that it's not a great idea just to have tactical, um, works in progress in an organization for ai, which could be way more than generative ai. They think it has to be across the board, everything full force, or it's no go.

I think that's horrible advice. I think that finding out what you want, what your problem statement is, what you're gonna solve with any type of ai, any, what are the projects that we could bring AI into? I think doing that tactically, one, one couple of pilot projects across different, um, domains in the organization will help you see, number one, what state is your data in?

Because if you're also, even with generative, if that data doesn't exist or it's not in the right form to be absorbed and, um, taken in by, uh, generative model, anything you get out of, it's not gonna be the way it should be. So that we've got the training issue, the adoption issue, but also, I, I don't agree with their advice to go 100% into to trying ai. I think it needs to be tactical, um, very specific, uh, projects that you're trying to use AI to I improve your business processes.

Let me have another twist to this thing. I think that you're also starting to see some frustration among the CEOs who were early adopters and first movers of this, because it's not being picked up as quickly as they might have hoped. And so they put a lot of money into it.

And then secondarily, it's also starting to be clear that a lot of these things that we invested in early are table stakes and are not necessarily gonna be competitive advantages. 'cause everybody's gonna catch up on that capability really quickly. So, you know, you gotta figure out, well, what is it that AI is actually gonna enable me and to have a unique advantage of?

Or is it just, you know, I need to invest in this because I need to stay even with everybody else, or I just can't play. Yeah, you, you know, I'm gonna echo what you just said, Mike, and what Fred say. You know, what always comes back to with Americans, I in particular I think is adopting technology.

Something new comes along. And as Fred pointed out with email, for instance, and I think AI is bigger and more intimidating, and, and the, the, the storyline, the narrative that the American workforce has is that this is gonna replace you. Um, and I I'll take my own example.

I mean, I was convinced that it was gonna eviscerate the entire journalism industry. It may, it may do eventually. Um, but the, the thing is, you, there's a a certain amount of, uh, skepticism of weariness, of using something without specific guidelines from the top management.

And, and the irony is, once you start using it, you realize what the benefits are. And it's, and it's a learning process. It's just not gonna happen overnight.

And I think there's a lot of impatience. I think that's the one word that always comes to my to mind about AI when I'm talking to these companies. And we were talking to the president of ServiceNow a couple days ago, and, and we push for examples.

They do. They are coming out, but they're not as, they're not, they're not a fusel lot of, of what you would expect. Um, Jensen mentioned Eli Lilly and you mentioned radiology and, uh, nursing industry, et cetera.

So they're, they are coming up with examples. But these things take time. And again, there's no patience in terms of this AI wave.

There's almost a sense of a panic of a gold rush still. Hey, if you watched the pit gen AI was on the pit. I don't know if you saw that, but I don't want to kill the episode per se.

But they were talking about how the physicians were gonna use that to generate their reports and might save some extra time. Do have, yes, of course. They also show that the Gen AI made an error and you, they have to carefully proofread everything that that gen AI creates.

Well, it's a great show. It's a great show. I I re-watched episode two of this last second season last night, actually, 'cause I couldn't find anything else on tv.

Um, guys, Fred, I thought you wanted to say something. I, I'm not sure. I would just add one last thing.

The way that we think about adoption of anything in business, uh, at, at least in my experience, starts with a use case. Gina mentioned this. When we look at the expansive improvement of an overall business using AI Echo, the patience and the conversation really has to be, which aspects of my bottom line am I seeking to improve?

What are the outcomes that I use to measure? We already see returns on investment for things like marketing and customer support and things like this. O obviously for development.

And, you know, whether or not we need to offshore something versus use ai, uh, and, and, and the gentech workflows for those work functions, those have to bear fruit over time. And those processes weren't built in a day to begin with before a ai. So translating that, I think also is gonna come into play with some of the things we're gonna talk about here in a minute or two around context and what context is required to make better decisions.

And some of those things have other limitations, like almost lot. So I think there's a lot of, uh, interesting movement, but patience is key. I mean, that's a Guns and Roses song that everybody should listen to.

Great enough. All right, let's jump to our next segment today then. And this one, well, it, it involves any ai, it involves AI as well.

It, you know, absolutely. It's around what's going on in the global memory market, memory chip market, actually, not the global memory market. That's pretty bad.

But memory chips on the other hand, no, that's not so great either. Mike, what's up here? Well, apparently the appetite for these things among the AI set is so high that is succeeding the capacity of the manufacturers of these memory devices and SSDs and everything that goes with that, to the point now where there's a shortage.

And we may not have enough of these things, not only to drive ai, but also a lot of our other IT projects and the cost of infrastructures going and go up. And there's a general feeling that the entire supply chain, at least as it relates to infrastructure, is becoming more and more constrained. And we may have to wait another year or so for manufacturing to catch up.

And that may reduce or limit the number of IT projects that get launched because simply the cost is going up and I might not even be able to get the things I need. Gina, we've been dealing with a GPU shortage for a while, but it looks like things might be getting a little worse. What do you think?

Uh, it's been bad for a little while. I think, honestly. So the, the thing to remember about GPUs is it's not the only way to accelerate compute.

You can also use memory. So Dr and Nan Flash have been around for a long while and people are using these to, especially in, um, the high performance computer area, which now is probably what we call ai. I don't know.

But, um, it's, it's been one of the problems I think, is that they're starting to build smaller and smaller types of memory and they're having problems doing that. So we talked about that I think last week with Intel. And there's also been with, um, with Samsung.

So the, the big three memory is Samsung Micron and SK Hynek, SK iex, I can't even say it. Um, but Samsung's had a problem with theirs. I know the plant that we have outside of Austin here in Taylor, Texas is not online yet.

I think it's been finished for a little while, not online yet, because Samsung can't get, they're having problems with the manufacturing of that smaller footprint, um, of, of, of the memory. And so, um, what's happened is, even though this is a really predictable memory, like we know what the supply chain is and we know when to get it, it's gone through the roof as far as prices go, is people try to get their hands on it as a backup to, to not being able to have their hands on GPUs. And so the same thing is, is going on with that.

And, um, you don't have manufacturers able to produce the new ones and or there's a slowdown in it is not gonna happen until next year, probably maybe the end of this year, but most likely not till 2027. So people are having to, um, either go it out or they've, or paid a high price for it. I know I've talked to customers, that's one thing they're looking at is how do they figure out how to make this run in a more efficient way so they don't need to rely on the memory since they can't get their hands on it.

So they're looking at, at things like virtualization, even deeper for ai. How does that work? Which, um, is a, a great thing to hear because we should be moving to better improvement with things like virtualization.

But if you look at like your cell phone or cars, I would think televisions, anything else that, that needs a membership to run, you probably are gonna, if you have to buy a new one, you're probably gonna be shelling out a little extra because they don't have the components they need or the prices for the components have gone way up. Uh, Fred, every time we run into a hardware issue around availability, somebody writes better software. So will this just be the mother as the necessity of invention here?

Absolutely. I also think it's, so it's, there's two forces at work. One, you know, the propensity to provide the supply chain with the appropriate amount of, of technology to support use cases and use cases.

And, and these use cases also include how do I include more data in context if I don't have to move, uh, you know, compute to the chip, uh, to the CPU, right? And I can compute in memory, then it not only saves an awful lot of resources and tokens, if I'm asking questions of a particular model or a, you know, a a a Frontier Labs, uh, service, uh, it, it dramatically changes the financial context that I'm gonna look at for using, you know, AI use cases. It also means that some of the requirements for GPUs can be offloaded in this sense.

And when I want to ask more intelligent questions, right? To our, to our, our segue from last, uh, episode is basically the conversation to suggest I can load up more of my valuable information into a memory which not only saves that cost on, on, uh, on token deployment, but it also means that I have better inference that is possible for a better answer in all cases, and I only have to load it once. So there's a huge value as people start to get more, uh, competent and utilizing prompting and agents to be able to load and access and have all those agents have access without having to, you know, go to compute in order to figure it out.

So the value is substantial and able to offload that to, you know, to memory chips. And I think that's, you know, just like we see the difference between caching stuff and memory or writing it to disk or a database, right? We're seeing that same value here, transitive value here of understanding why that's important for something like memory versus, you know, GPU.

So I, I think absolutely, Mike, it's a, it's gonna be a, uh, uh, a, uh, what do we used to call those things that would go back and forth on the playground? Teeter-totter. Teeter totter.

You, Seesaw, Seesaw With the nails, Alan, Alan, are Americans gonna be willing to pay more for a television so Fred can cash more data in his ai? What do you think? Well, yeah, we don't care what those TVs cost, but, but I'm glad you framed it that way because there's an issue here and that issue's digital sovereignty.

Mm-hmm. Right? Mm-hmm.

Digital sovereignty was kind of the topic at Devo Davos this week. Well, maybe Iceland or Greenland was, but digital sovereignty's very real. It's very real.

And you know, even sort of not as cutting edge chips like these high performance memory chips compared to GPUs, there's a very finite market and there's a very finite, finite geography of the players who can produce these kinds of chips, let alone the GPUs and the, and the inference ships and everything else. Now, on my predict 2026, uh, session I did with Daniel Newman, we spoke about, you know, he, he doesn't think there's any AI bubble. I think there's a financial bubble.

And his reasoning is that right now the demand is for 175% of what we are producing in ways, in terms of semis, in terms of chips, in terms of hardware, silicon. And as long as that demand far exceeds the ability to produce, you're gonna see prices, you're gonna see stock market prices reflect market pricing, reflect that as well. We get a lot more surveys like we discussed in, in, uh, the first segment.

And maybe that demand falls down and you will see a financial air come out of this balloon. But let me return to sovereignty. If you're in Europe and these new plants coming online this year, Gina are in Tyler, Texas, I'm not jumping up and down saying, Yee high, can't wait for that plant to open up.

You're thinking to yourself, I am not going to be renting an apartment from a crazy landlord, and I I'm gonna look for an apartment near my, where I live now that's a little bit more stable and a little bit less prone to, you know, controversy or, or it's instability. And so that's what's going on here, right? How many of these memory chips are gonna be built in Korea are gonna be built in Taiwan, they're gonna be built in China versus where they're gonna be used, right?

Samsung's doing this plant in Tyler if they get the, the technology right? If they, you know, one, one thing about the American chip manufacturing business is we have not proven the ability yet to have enough of a trained workforce and, and the technolo technological chops to produce the cutting edge chips, though Intel did make that announcement a couple weeks ago that they are on track with their latest kind of, I think it was two nano I, or I forgot the designation. So, you know, whether or not we can produce domestically the chips we need to, to, you know, do what we want to do is still an open question.

And sovereignty, digital sovereignty is going to play its role here, Right? Without making a nod to maybe David Stockman and supply side economics. But, um, if I, if the cost of a airplane flight continues to rise, fewer people get on the plane and travel as far, and they then generally drive somewhere else and locally and whatever their economics can afford.

And the same thinking is gonna apply to AI here. People are just gonna say, uh, you know what, uh, I wouldn't have loved to fund 10 projects, but we can only afford five. That's all there is to it, right?

Yeah, That's, that's interesting you said, it's not that this affordability issue doesn't apply just to the consumers, but we mentioned the data centers and the ripple effect that might have on some of these companies building out. 4 trillion in, in costs? God, it, it, the math to me, it always keeps coming back to the math and it just doesn't add up.

And this just, it, it adds pressure to the pressure cooker, especially for a company like OpenAI. And it just, it's just not sustainable. And to that s and p report that you were pointing out too, I mean, at its core, when it's pointing to is the first wave of these data centers were funded by the hyperscalers and the METAS and all those people, but now the next wave, they're looking to things like the bond market and investors to fund, and that's much more speculative.

And some of those people apparently according to SAP, like, are already pulling back wondering about, well, whether there's gonna be an ROI on it for them. And so it's getting harder to find the cash to build these data centers S and their money. I think one more thing, just real quick is, um, about the Samsung kind.

I know one of the other issues is there's one being built in Georgia as well, and that big fusel kerfuffle with integration has the go Georgia governor back and forth to Korea trying to smooth things over to make sure that plant actually gets completed. So we're running into things where the geopolitical is, um, really impeding on our ability to create manufacturing plants in the us. So that adds to the issue as well.

Yeah. Not just in the US overseas we call an $8 trillion of, uh, bonds and, you know, that'll crater any economy. So, you know, the, the geopolitical landscape here dramatically affects where you going to build data centers who, you know, shimmy the data sovereignty problem that we see in data and transit and data and storage from a privacy perspective now, right?

All the way down to who is going to own and operate, you know, these data centers that are gonna have the impact And and where the, where the, where the equipment for those data centers come from. Exactly. Are they gonna have back doors?

Are they gonna be controls? Can I get replacement parts, right? AI sovereignty.

I don't want to use A-A-A-A-A model that was developed and based in, in one country versus another. I want my own AI model. You mean I can't make memory chips with my 3D printer in the corner?

What are you saying? That'd be awesome. You know, this is, what do you need memory chips for?

But you know, this is the world we live in. But Mike, you said something and I'll end this segment on that. Necessity is the mother of invention.

And one thing that the tech industry is, and entrepreneurs are, are resourceful, innovative, right? And resilient. And where there's a will, there's a way they'll figure it out.

Let's jump to our next section and leave it on a positive note like that. Um, elephants for no water free, water free. I thought it was elephants water.

It's water free data centers. What do we got, Mike? That's a perfect segue.

So Microsoft is talking about how they are gonna build a data center largely for these AI types applications that doesn't require water to cool all the equipment, which, you know, is a big issue if you've been following this whole data center environmental impact debate. Uh, it's not just the noise, it's the amount of water that gets consumed by these things and how that drives up local, uh, or limits the amount of resources people have locally. And maybe there's also conversations about, well, what happens to that water when it, when they're done with it.

But, um, it seems like maybe we can innovate our way out of this thing. So Alan, I know you've been a big proponent of this notion, but, you know, is this something that gives you hope? Yeah, I mean, you know, it's interesting they're doing this in West Des Moines, but I don't know if they're doing it in East Des Moines.

I didn't realize Des Moines was big enough to have a west and an east. But that, that being said, the, the fact of the matter is, you're right. I mean, the, the amount of water that would be required to cool down, you know, this 8 trillion data center build out really doesn't exist.

I mean, I guess you could bring seawater in if it wouldn't corro corrode everything. But, but here's the other thing is from what I've been told, and I'm no expert on this of course, but I did stay in a holiday in express last night. The the water doesn't cool it down enough.

You need liquid cooling. Air cooling doesn't work at these levels when you've got, you know, gigawatt racks and so forth, air cooling doesn't work. You need liquid cooling, but the, you know, just plain old water isn't efficient enough in cooling it down.

This is akin to putting water in your radiator of your car versus antifreeze, right? And the antifreeze isn't just about not freezing, it also protects it from running hot, right? Because water boils at 180 and you got steam and hey, that's good.

If you want to convert the water you use in your data center into heating, uh, steam for, for, you know, cold weather climates like in Des Moines, they're in the winter, but you know, they, kudos to Microsoft and, and actually the companies behind this, the, these innovations. But are we replacing water with antifreeze? Is it, you know, is that the, is that the question here?

And um, you know, that antifreeze is pretty toxic. Is this stuff going to be toxic? How are we get rid of it?

That's what I thought immediately. So especially from the headlines, right? Like zero water data center, what the heck does that mean?

But, um, it is pretty cool the technology that they're using, because they are doing it specifically, they're, they're able to, um, to take the, the water, the, the, the liquid cooling, which right, they don't anti Antifreeze, It's antifreeze or it's, it doesn't sound like it's antifreeze unless this is what antifreeze is. It has like silicone or something in the water, right? So they're are mineral oil is what they put in the water.

So they, they, they, they take that they can deliver it specifically to the component that's overheating. And it, what it is, is it's a little, there's little, uh, plates that fit on top of the components and they can go right to the one that's overheating that plate will get all of it, will take all of the heat off of it using that mineral water liquid, whatever it is, and take it back to something else that will then cool it down. And they can just redistribute that continuously through the system.

Um, which I think is like brilliant. So if you know something's hot, you get to send it to that. So, but my question is, you know, they're, they're able to cool this without the water.

They're able to do it specifically to what it, the, what needs to be cooled down. So that helps with the electricity as well. But you know, what, how toxic is it?

What's the danger of that bursting in in something to, you know, having some kind of e ecological issue, but also is this just heat sinks? I just think it's super charged heat sinks, but I'm not sure that's what it sounded like to me. So can I just take the antifreeze and bury 'em in that soon to be depleted coal mine?

Because you know, we got all that space, we might as well just think, Hey, you know, you can, you could use bleach in a UV light to cure some diseases too, right? But, um, but, but, you know, but, but let's be real here, right? It's an efficiency over what we have now.

And sometimes, you know, I do, a lesson I learned from Brad Feld is things aren't always revolutionary. They're evolutionary and we take little steps, little steps. And when you look back over the arc of time, those little steps make for giant leaps, right?

Someone said that about landing on the moon or something, right? Giant leaps. And so, you know, this is, whether it's a little step or a medium step, it's a step in the right direction.

And, and so it's Weird in a sense, Microsoft is trying, is kind of taking a lead in, in kind of being responsible about their data centers. Maybe this is just good PR on their part, but it, not just with water, but with energy costs and assuming some of those costs. So I, you know, hopefully other companies follow suit and try to go down this environmental path.

I know Nvidia always plays up its green approach, but, um, you know, I like to see a little bit more practical or pragmatic, uh, solutions from them. Like Microsoft is providing. You're, you're, you're right about that, John.

They got the nuclear research at Three Mile Island. They're also leaders in solar. I mean, they're probably staying away from windmills just to prevent themselves from To a piece, somebody Yeah.

You know, generally speaking. But it's a cool, I thought it was pretty cool. Um, but it does kind of remind me of just like, you know, next gen heat sinks just like really cool because if, um, if they can do it for, for servers, a lot of that, a lot of the heat needs to be resolved in consumer products too, like games.

So, or my laptop in particular, which has an end Nvidia chip. So, but like, it would be pretty cool if you just, you know, like how do you provide that amount of cooling to, uh, consumer products? Um, and can you do it with this, I'm calling it next gen heat sink, but I'm not, I don't know what it is, but that's what it sounded like to me.

Like, how do you put something on top of a chip and then you, I I'm sure they have some monitoring that goes with it. The whole idea is fascinating to me. Mm-hmm.

So, go ahead. If we save 125, uh, million liters of water per day and we consume on average, uh, 5 million gallons of water per data center, uh, in this stretch, and Microsoft is gonna build, I think it was 287 data centers for AI generation, uh, i, I would say it's more than altruism, right? To optimize the package, uh, the value I think is obviously going to be an incremental value.

But if the usage is going to double or quadruple by 2028, this is more than an efficiency gain in a right now problem. And I think, you know, before we get to the quantum photonics chips, this is the only game in town. So savvy to do it now Absolutely.

Is That, and I think Bill continue to improve it. I'm sorry, go ahead, Mike. Is That the same math my kids use?

'cause they come home and they'll say, you know, I saved a hundred dollars on this $200 thing that I bought. So I go, well, does that mean when I bring it back I get $300? 'cause that doesn't usually work out that much.

I totally agree that math always stands the touch of time. You Know, it's, they say in Las Vegas at the gaming tables. It's a fine beginning.

All right. Hey, we're at our end though. So instead of a beginning, we're gonna end our text on gang today.

Gina, Fred, John, Mike, thanks for joining. Thank you for watching. We love doing this live.

We'd love to hear from you. So do, you know, put your comments in and what have you. Of course, if you're watching this on the, uh, text drum TV stream after, thank you for watching it there, or wherever you're watching this on YouTube, our text drum tv, YouTube channel text strong tv OTT app on iOS and Android and Roku and Apple TV and, uh, Amazon Fire, or on the Textron TV website or any of the Textron sites for that matter.

Thank you. We appreciate it. Hey, we've got Text Strong, uh, TV immediately following this, and at two 30 Eastern time today, I'll be live on, uh, this week.

Shimmy says, this week, shimmy says, is DevOps the never ending story? Um, so check that out. I'm also doing a shimmy says tomorrow.

Check that one out as well. That'll be on digital sovereignty, and we'll talk more about some of the things we spoke about here. But for now, this is Alan Shimel.

We're out.